Abstract: To improve the medical image fusion effect, a novel method with region-based direction feature is proposed with wavelet transform. The direction information of the source images is extracted by the eigen-values and eigen-vectors of structure tensor of the every source image. The eigen-values are used to devise an index to show the local feature of the image. The index is averaged to yield the weight for adaptive high frequency coefficient fusion. We apply the proposed approach to the fusion of medical images and the experimental results show the performance of the new method.